首页|期刊导航|中国医学工程|人工智能在pT1期结直肠癌肿瘤形态定量评估中的研究进展

人工智能在pT1期结直肠癌肿瘤形态定量评估中的研究进展OA

Research progress of artificial intelligence in quantitative analysis of tumor morphology in pT1 colorectal cancer

中文摘要英文摘要

近年来,人工智能在病理学领域展现出巨大的应用潜力.结直肠癌是全球发病率第三的恶性肿瘤,其相关性死亡率位居第二,已成为重大公共卫生问题.pT1期结直肠癌指肿瘤浸润至黏膜下层但未累及固有肌层,淋巴结转移(LNM)是影响其术后复发和死亡的重要原因,准确评估LNM风险对制定个体化治疗方案至关重要.目前临床主要依赖淋巴血管侵犯(LVI)等定性特征进行预测,而对肿瘤形态定量参数(如浸润深度、面积等)的评估仍较为欠缺.随着数字病理与人工智能技术的发展,基于全切片数字图像(WSI)和深度学习算法的病理大模型为结直肠癌淋巴结转移的精准预测提供了新方法.本文通过计算机检索与手工检索相结合的方式,系统梳理国内外近20年基于病理大模型的pT1期结直肠癌形态定量分析与LNM相关性的研究文献,围绕人工智能(AI)在结直肠癌病理图像中的自动化分析、形态定量评估、多模态数据整合等方面进行综述,结合各研究的试验设计、研究人群、技术平台等背景差异,对比分析不同研究结果的异同及成因,并深入探讨当前领域存在的技术不足,为AI在pT1期结直肠癌病理诊断中的临床转化提供参考.

In recent years,artificial intelligence has demonstrated significant application potential in the field of pathology.Colorectal cancer is the third most common malignancy globally and ranks second in cancer-related mortality,making it a major public health concern.pT1 stage colorectal cancer refers to tumors that invade the submucosa but do not involve the muscularis propria.Lymph node metastasis(LNM)is a critical factor contributing to postoperative recurrence and mortality in these patients,making accurate LNM risk assessment essential for personalized treatment decisions.Current clinical practice primarily relies on qualitative features such as lymphovascular invasion(LVI)for prediction,while quantitative morphological parameters(e.g.,depth of invasion,tumor area)remain underexplored.With the advancement of digital pathology and artificial intelligence,whole slide image(WSI)-based pathological foundation models utilizing deep learning algorithms offer a novel approach for accurately predicting lymph node metastasis in colorectal cancer.This review systematically examines studies from the past 20 years on the correlation between morphological quantitative analysis and LNM in pT1 colorectal cancer using pathological foundation models,retrieved through both computerized and manual searches.The discussion focuses on AI-driven automated analysis of pathological images,quantitative morphological assessment,and multimodal data integration.Considering variations in study design,study populations,and technological platforms,we compare and analyze similarities and differences in research findings and explore their underlying causes.Furthermore,we delve into the existing challenges,particularly the lack of technical standardization,to provide insights for the clinical translation of AI in pT1 colorectal cancer pathological diagnosis.

陈霜;吴进督;高银华;刘汉忠

武汉科技大学附属孝感医院(孝感市中心医院),湖北 孝感 432000||武汉科技大学,湖北 武汉 430081武汉科技大学附属孝感医院(孝感市中心医院),湖北 孝感 432000||武汉科技大学,湖北 武汉 430081武汉科技大学附属孝感医院(孝感市中心医院),湖北 孝感 432000||武汉科技大学,湖北 武汉 430081武汉科技大学附属孝感医院(孝感市中心医院),湖北 孝感 432000||武汉科技大学,湖北 武汉 430081

医药卫生

结直肠癌病理诊断人工智能定量分析

colorectal cancerpathological diagnosisartificial intelligencequantitative analysis

《中国医学工程》 2026 (8)

106-110,5

孝感市中心医院2026年度院级课题(XGZX202602,重点项目)

10.19338/j.issn.1672-2019.2026.08.019

评论